5 ms·

A Cartographer Who’s Transforming Map Design

131 points by kurren 11y ago

wscott 11y ago
The "Tufte" of maps.
rubidium 11y ago
Some of her earlier papers can be found here: http://www.personal.psu.edu/cab38/Pub_scans/Brewer_pubs.html http://www.personal.psu.edu/cab38/Pub_scans/Brewer_pubs.html Here's the ColorBrewer tool: http://colorbrewer2.org/ http://colorbrewer2.org/ I never thought about what a PhD in cartography would look like. Nice to learn more about the thought being put into maps.
- pella 11y ago
  more presentation: http://www.personal.psu.edu/cab38/ScaleMaster/ http://www.personal.psu.edu/cab38/ScaleMaster/
ipunchghosts 11y ago
I sit a few buildings down from Cindy's office and have also been studying colormaps quite a bit in the last 3 years. Its interesting how many plotting packages get this wrong but are finally catching up. I switched from Matlab to Python years ago and was sad to see pyplot using the default rainbow palette still. However, there was some good work done by Chris Beaumont to improve the plot quality. See: http://plotornot.chrisbeaumont.org/ http://plotornot.chrisbeaumont.org/ You can easily import these styles into matplotlib using rcparams. Matlab is using a roughly perceptually linearly luminant colormap they call Parula now. Good job Matlab. Paulo Penteado has also done some good work in this area. See: http://www.ppenteado.net/ast/csbc2012_pfp_2_pres.pdf http://www.ppenteado.net/ast/csbc2012_pfp_2_pres.pdf I want to talk about the Luv Lab colorspace. There are several places on the net (even in the literature) that are wrong about these colorspaces saying Lab is for emissive displays and Luv is for reflected light. This is actually not true. (If anything it is reversed). See: https://groups.google.com/d/msg/scikit-image/DIRaSXJoEes/2jDwuwmxRTYJ https://groups.google.com/d/msg/scikit-image/DIRaSXJoEes/2jD... and Berns reference. The interesting with colorspaces (and colormaps thereof) is that working in a perceptual space like Luv/Lab is yields a non-linear (and non-convex) gamut in the sRGB space used by most monitors. There is more "headroom" in the magenta hue of colors than say green. However, you have to then look at monitor output as a function of hue and human sensitivity -- with a red object and blue object with the same reflectance under the same illumination, the red object will appear darker to humans. So there are many transfer functions at work here which makes the problem challenging in picking the right colormap that is perceptually uniform, has the maximum number of perceived differences, and has the appropriate number of hues for best represeting your dataset. Finally, it would not be complete of me to not mention this article: http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=4118486&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D4118486 http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=411848... Slides: http://www.cs.odu.edu/~mweigle/cs725s15/presentations/nam-presentation.pdf http://www.cs.odu.edu/~mweigle/cs725s15/presentations/nam-pr...
- urschrei 11y ago
  > Matlab is using a roughly perceptually linearly luminant colormap they call Parula now. Good job Matlab. And Matplotlib is switching to a perceptually-linear default colormap called "viridis" with the 2.0 release (due early this month): http://matplotlib.org/style_changes.html http://matplotlib.org/style_changes.html
  - jacobolus 11y ago
    I’m glad that people are finally catching up to Cynthia Brewer (who has been discussing the issue for 20+ years) and using reasonable color models to think about data displays. It would be nice if color pickers, color manipulation tools in graphics software, etc. would also switch to more perceptually relevant models. If anyone wants to play with the CIECAM02 model discussed in the talk you linked, I made an implementation in Javascript several years back, https://github.com/jrus/chromatist/ https://github.com/jrus/chromatist/
- muxxa 11y ago
  For perceptual color theory another good one is the Munsell palette which has been around for a while: munsell.com
- rrherr 11y ago
  For Color Brewer palettes in Python, you can also use Seaborn, which is based on matplotlib. http://stanford.edu/~mwaskom/software/seaborn/tutorial/color_palettes.html http://stanford.edu/~mwaskom/software/seaborn/tutorial/color...
gavreh 11y ago
Here's the blog post about how her work has influenced Esri's ArcGIS software: http://blogs.esri.com/esri/arcgis/2014/11/12/brewing-a-new-color-palette-for-arcgis-pro/ http://blogs.esri.com/esri/arcgis/2014/11/12/brewing-a-new-c...
- anc84 11y ago
  Which the FLOSS QGIS natively had available for a long time before them. ;)
larrydag 11y ago
Here is the popular Color Brewer implementation in R. https://cran.r-project.org/web/packages/RColorBrewer/index.html https://cran.r-project.org/web/packages/RColorBrewer/index.h... A tutorial on how to implement RColorBrewer in R. http://www.compbiome.com/2010/12/r-using-rcolorbrewer-to-colour-your.html http://www.compbiome.com/2010/12/r-using-rcolorbrewer-to-col...
electricblue 11y ago
I've been using colorbrewer to make maps for many years now, nice to see her getting some recognition.
hackuser 11y ago
The article is dated, very subtly, 2014. Look just above the headline, next to the byline.
heuermh 11y ago
This java library (among several others, certainly) provides color schemes for data viz, including the ColorBrewer schemes http://www.dishevelled.org/color-scheme/ http://www.dishevelled.org/color-scheme/

131 points by kurren 11y ago

The "Tufte" of maps.

Some of her earlier papers can be found here: http://www.personal.psu.edu/cab38/Pub_scans/Brewer_pubs.html http://www.personal.psu.edu/cab38/Pub_scans/Brewer_pubs.html Here's the ColorBrewer tool: http://colorbrewer2.org/ http://colorbrewer2.org/ I never thought about what a PhD in cartography would look like. Nice to learn more about the thought being put into maps.

pella 11y ago
more presentation: http://www.personal.psu.edu/cab38/ScaleMaster/ http://www.personal.psu.edu/cab38/ScaleMaster/

ipunchghosts 11y ago

I sit a few buildings down from Cindy's office and have also been studying colormaps quite a bit in the last 3 years. Its interesting how many plotting packages get this wrong but are finally catching up. I switched from Matlab to Python years ago and was sad to see pyplot using the default rainbow palette still. However, there was some good work done by Chris Beaumont to improve the plot quality. See: http://plotornot.chrisbeaumont.org/ http://plotornot.chrisbeaumont.org/ You can easily import these styles into matplotlib using rcparams. Matlab is using a roughly perceptually linearly luminant colormap they call Parula now. Good job Matlab. Paulo Penteado has also done some good work in this area. See: http://www.ppenteado.net/ast/csbc2012_pfp_2_pres.pdf http://www.ppenteado.net/ast/csbc2012_pfp_2_pres.pdf I want to talk about the Luv Lab colorspace. There are several places on the net (even in the literature) that are wrong about these colorspaces saying Lab is for emissive displays and Luv is for reflected light. This is actually not true. (If anything it is reversed). See: https://groups.google.com/d/msg/scikit-image/DIRaSXJoEes/2jDwuwmxRTYJ https://groups.google.com/d/msg/scikit-image/DIRaSXJoEes/2jD... and Berns reference. The interesting with colorspaces (and colormaps thereof) is that working in a perceptual space like Luv/Lab is yields a non-linear (and non-convex) gamut in the sRGB space used by most monitors. There is more "headroom" in the magenta hue of colors than say green. However, you have to then look at monitor output as a function of hue and human sensitivity -- with a red object and blue object with the same reflectance under the same illumination, the red object will appear darker to humans. So there are many transfer functions at work here which makes the problem challenging in picking the right colormap that is perceptually uniform, has the maximum number of perceived differences, and has the appropriate number of hues for best represeting your dataset. Finally, it would not be complete of me to not mention this article: http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=4118486&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D4118486 http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=411848... Slides: http://www.cs.odu.edu/~mweigle/cs725s15/presentations/nam-presentation.pdf http://www.cs.odu.edu/~mweigle/cs725s15/presentations/nam-pr...

urschrei 11y ago
> Matlab is using a roughly perceptually linearly luminant colormap they call Parula now. Good job Matlab. And Matplotlib is switching to a perceptually-linear default colormap called "viridis" with the 2.0 release (due early this month): http://matplotlib.org/style_changes.html http://matplotlib.org/style_changes.html
- jacobolus 11y ago
  I’m glad that people are finally catching up to Cynthia Brewer (who has been discussing the issue for 20+ years) and using reasonable color models to think about data displays. It would be nice if color pickers, color manipulation tools in graphics software, etc. would also switch to more perceptually relevant models. If anyone wants to play with the CIECAM02 model discussed in the talk you linked, I made an implementation in Javascript several years back, https://github.com/jrus/chromatist/ https://github.com/jrus/chromatist/
muxxa 11y ago
For perceptual color theory another good one is the Munsell palette which has been around for a while: munsell.com
rrherr 11y ago
For Color Brewer palettes in Python, you can also use Seaborn, which is based on matplotlib. http://stanford.edu/~mwaskom/software/seaborn/tutorial/color_palettes.html http://stanford.edu/~mwaskom/software/seaborn/tutorial/color...

gavreh 11y ago

Here's the blog post about how her work has influenced Esri's ArcGIS software: http://blogs.esri.com/esri/arcgis/2014/11/12/brewing-a-new-color-palette-for-arcgis-pro/ http://blogs.esri.com/esri/arcgis/2014/11/12/brewing-a-new-c...

anc84 11y ago
Which the FLOSS QGIS natively had available for a long time before them. ;)

larrydag 11y ago

Here is the popular Color Brewer implementation in R. https://cran.r-project.org/web/packages/RColorBrewer/index.html https://cran.r-project.org/web/packages/RColorBrewer/index.h... A tutorial on how to implement RColorBrewer in R. http://www.compbiome.com/2010/12/r-using-rcolorbrewer-to-colour-your.html http://www.compbiome.com/2010/12/r-using-rcolorbrewer-to-col...

electricblue 11y ago

I've been using colorbrewer to make maps for many years now, nice to see her getting some recognition.

hackuser 11y ago

The article is dated, very subtly, 2014. Look just above the headline, next to the byline.

heuermh 11y ago

This java library (among several others, certainly) provides color schemes for data viz, including the ColorBrewer schemes http://www.dishevelled.org/color-scheme/ http://www.dishevelled.org/color-scheme/